CAUSAL-MODELING TO ESTIMATE SENSITIVITY AND SPECIFICITY OF A TEST WHEN PREVALENCE CHANGES

The objective of this paper is to demonstrate, by causal modeling, whe ther the sensitivity and specificity of a test are constant, or whethe r they change with prevalence. I use three assumptions, diagnostic, pr edictive, and correlational, in three sets of mathematical models. A d iagnostic test measures an outcome of a disease and is based on the as sumption that the ''gold standard'' result indicates a disease state t hat causes a test result. A predictive test measures a risk factor for a disease and is based on the assumption that the test result indicat es a risk factor that causes a disease state. A correlational test mea sures a condition which is an outcome of an underlying causal risk fac tor for a disease and is based on the assumption that the disease and the test result are noncausally related. I find that sensitivity and s pecificity are constant for diagnostic tests but change with prevalenc e for predictive and correlational tests. I present equations to show the effects on various test performance indices when prevalence change s. Different equations must be used to estimate the sensitivity, speci ficity, and other test performance indicators for various types of tes ts that are under different causal assumptions.